How our predictions work

The percentages describe the market estimate for a home win, draw and away win in regulation time.

How probabilities are calculated

We take 1 / decimal odds for each outcome, then divide each value by the sum of all three. This removes the margin from that line and produces probabilities that add up to 100%. When several fresh lines are available, we average their normalized probabilities.

Sources and updates

API-Football supplies schedules, scores, standings, lineups and events. Probabilities use bookmaker odds as a statistical reference. These data are independent of affiliate offers and do not represent our partners’ prices. We only use lines updated within the last two hours. Best prices are compared for each outcome; they are not presented as an independent model. English kickoff times and match URL dates use America/New_York, including daylight saving time.

Recent form and head-to-head

Form uses up to five completed matches with known scores for 90 minutes plus stoppage time, excluding extra time and penalties. Historical previews only include results from before kickoff. Head-to-head records show up to five previous meetings. These records provide context and do not change the probabilities calculated from odds.

GidStats opinion

This section gives an outcome preference, supporting evidence and a specific risk, such as an opponent’s draws or goals conceded by the favorite. It is generated automatically from results available before kickoff. It compares home and away performance first, then overall form if there is no clear difference. The preference must be supported by fresh market odds, a saved pre-match calculation or a points-per-game advantage in the same comparable league table. Goals use known regulation-time scores. The section is omitted when data are insufficient or do not support either team. It adds no new probabilities to the market estimate and does not imply that each match has been manually reviewed.

Measuring accuracy

Only probabilities saved before kickoff are evaluated against regulation-time results. Predictions are never reconstructed after the result. The report shows the sample size, the hit rate of the most likely outcome and the Brier score.

Limitations

A high probability does not guarantee an outcome. Data can arrive late, and some competitions do not provide lineups. This information does not promise a profit.